Walking in the dark

[Avery Andrews 931126.1453]

A few points on closed/open loop:

  1) Contrary to what I said a week or so ago, I now see walking in the
      dark (for ordinary people, not practiced wall-hearers) as
      open-loop w.r.t. *location* in the room. It is of course
      closed-loop w.r.t. many other things (kinesthesia, acceleration
      (via the vestibular organs). But, since there is no interaction
      with the distal features of the room whereby location in the room
      is defined, it is open-loop w.r.t. location.

  2) But, as Hans emphasizes, there is an inner model that produces a
      felt sense of location, based on information about initial
      position, and integration (in both intuitive and mathematical
      senses) of information from the other available perceptions
      (and maybe efferent-copy as well, at least in principle).

  3) The model updating is not imagination - one is not predicting
      hypothetical location on the basis of hypothetical locomotion,
      but *feeling* (perhaps incorrect, but putatively actual) location
      on the basis of (ordinary, closed-loop regulated) perceptions of
      motion and acceleration.

  4) The required kind of model-updating might be a pre-adaptation
      to full-scale imagination.

  5) Calling model-based control `open-loop' is a good way to draw
      attention to its inherent limitations (the less predictable the
      environment, the shorter the time-span for which it works well
      enough to be useful).

  6) We need some new terminology. `perception' is often taken to
      imply conscious awareness, which is not implied in PCT. There
      should also be terminology to distinguish between
      sensing/perception via actual interaction, vs. modelling.
      The sense/model and aware/unaware distinctions give is four
      concepts, not including imagination.

  7) Modelling never works forever: the model has to be kept in tune
      by comparision with an actual (interatively-based) perception,
      which involves at least: (a) resetting the value to whatever is
      actually observed (when the lights go on, you *see* where you are,
      and the old position-sense based on dead-reckoning is erased by
      the new one based on vision). (b) `tuning' (changing the structure
      of the model to minimize the size of the necessary resettings).
      The model is `of' whatever CEV is used to reset and calibrate it
      (I wonder if there isn't a solution to some philosophical problems
      of aboutness of representations here).

  8) An everyday example of model-based `control' is the activity of
      staying with someone while walking around in a shopping center
      with them. You don't keep your eyes on them every instant, but
      assume that their behavior will observe certain regularities
      (such as turning corners necessary to get to places you both agree
      you will go). You have a sense of `being with' that person, which
      you are controlling for maintaining, but sometimes this sense
      becomes erroneous because the model generates wrong results
      (you think they turned the corner because that's how you think
      one gets to where you're going; they go straight ahead because
      that's what they think, and weren't watching you because they
      thought they knew where you were going, and were looking at
      stuff in a shop-window.


From Chris Malcolm

[Avery Andrews 931126.1453 writes:

1) Contrary to what I said a week or so ago, I now see walking in the
     dark (for ordinary people, not practiced wall-hearers) as
     open-loop w.r.t. *location* in the room.

You have to be careful judging this one. Some people find that they can
hear where walls etc. are. Some people are convinced that they can't,
but tests (e.g. asking them to guess where the walls are) show that they
actually can, but are not simply not aware of the perception, rather
like blind-sight. When placed in carefully contrived situations, some
people who thought they couldn't hear walls discover that they can;
others, whom tests show can do it, still remain stubbornly "blind" to
awareness of the perception. And people's behaviour can be influenced
by this kind of perception they are not aware of having.

I have a suspicion that some subtle and important truths about the
functional role of consciousness are lurking around here.

Chris Malcolm